gmiesw / gmiesw/example-repository
Model output is not the same after speedup.
Open
customer raised
good first issue
- Dominant language
- Apex
- Stars
- 0
- Forks
- 0
- PR merge metrics
- No merged PRs in 30d
Description
Environment:
- NNI version: commit f1b8cd2
- NNI mode (local|remote|pai): remote
- Client OS: Ubuntu 16.04.5 LTS (Xenial Xerus)
- Python version: Python 3.6.4
- PyTorch/TensorFlow version: PyTorch 1.5.0
- Is conda/virtualenv/venv used?: pyenv
- Is running in Docker?: No
Log message:
- nnimanager.log:
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.0.conv1
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.0.conv2
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.2.conv3,layer3.1.conv3,layer3.0.downsample.0,layer3.5.conv3,layer3.0.conv3,layer3.3.conv3,layer3.4.conv3
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.1.conv1
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.1.conv2
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.2.conv1
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.2.conv2
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.3.conv1
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.3.conv2
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.4.conv1
INFO:nni.compression.torch.pruning.one_shot:Pruning the dependent layers: layer3.4.conv2
What issue meet, what's expected?:
I have different outputs before and after model speedup:
It's expected that outputs would be roughly the same.
How to reproduce it?:
Please, run the code from the following gist:
https://gist.github.com/nikitakaraevv/1eaa735d5323f833356152f6d41ef8e4
Additional information:
I'd like to understand why it happens first of all and how it can be avoided.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.